A very similar method may be used to determine the atomic weights of the atoms of which the stars are composed. A star is in effect nothing but a huge X-ray apparatus. We know the weights of many of the stars, and the rate at which they are generating X-rays is merely the rate at which they are radiating energy away into space. If we could cut each atomic nucleus in a star into two halves, we should halve the opacity of the star, so that radiation would travel twice as far through the star before being absorbed. If the star were wholly gaseous, this would result in its expanding to four times its original diameter, and in its surface-temperature being halved. It follows that we can calculate the atomic weight of the atoms of which a star is composed from the weight, luminosity and surface-temperature of the star. The atomic weights of a number of stars, which I calculated on the supposition that the stars were wholly gaseous, came out in practically every case higher than that of uranium, which is the weightiest atom known on earth. They not only proved to be higher, but enormously higher; so high indeed, as to seem utterly improbable. Again the explanation seems to be that the stars are not wholly gaseous. As soon as stellar interiors are supposed to be partially liquid, the calculated atomic weights are reduced enormously. They can no longer be determined exactly, but the atomic number of about 95 to which we were led from a consideration of the Russell diagram seems to be entirely consistent with all the known facts. Indeed other considerations seem to suggest that the atomic numbers of stellar atoms must be higher than 92. _A priori_ stellar radiation might either originate in types of matter known to us on earth or else in other and unknown types. When once it is accepted that high temperature and density can do nothing to accelerate the generation of radiation by ordinary matter, it becomes clear that stellar radiation cannot originate in types of matter known to us on earth. Other types of matter must exist, and as, with two exceptions, all atomic numbers up to 92 (uranium) are already occupied by terrestrial elements, it seems probable that these other types must be elements of higher atomic weight than uranium. These super-heavy atoms must not be expected to disclose their presence in stellar spectra, for these only inform us as to the constitution of the atmospheres of the stars. And as the lighter atoms float to the top it is these, in the main, which figure in stellar spectra. If the sun’s atmosphere had contained any considerable number of super-heavy atoms when the planets were born, some of them ought still to exist in the earth. There cannot be any great number, or their high generation of energy would betray them. The simplest view seems to be that the heavier atoms sink to the centre in the stars, and that the earth was formed mainly or solely out of the lighter atoms which had floated to the sun’s surface.